A major challenge in the field of extragalactic astrophysics is understanding when the most massive galaxies form the bulk of their stars, and determining the specific pathways they take to assemble mass across a wide range of redshifts and environments. In this thesis, I describe my work to characterize stellar populations in massive galaxies at two epochs — redshifts ~0.5 and ~5. I use spectra of galaxies in massive South Pole Telescope galaxy clusters to address the question: on what timescales do galaxies that end up in clusters form their stars, and does the cluster sample matter when studying these properties? This mass-limited cluster sample across redshifts 0.3 < z < 1.5 allows me to constrain star formation histories and formation redshifts of 900 quiescent galaxies in clusters, as a function of cluster environment and mass. This study explores mass-dependent evolution in cluster quiescent galaxies and characterises galaxy evolution across a descendent-antecedent cluster sample. On the other ‘end’ of the redshift scale, I describe the characterization of COOL J1241+2219 (or CJ1241), a lensed galaxy at z = 5.04 that is the brightest galaxy known at z > 5 (at AB magnitude z ~ 20.5). This galaxy was discovered by COOL-LAMPS — ChicagO Optically-selected strong Lenses - Located At the Margins of Public Surveys — initiated to find strongly lensed systems, consisting primarily of a team of undergraduate students. I characterize the lensed galaxy using ground-based spectrophotometric data to find an intrinsically luminous and massive star-forming galaxy. In this thesis, I also show first results aimed at comparing CJ1241 and other COOL-LAMPS discovered lensed massive galaxies at z > 3 with their potential descendents — quiescent massive lensed galaxies at lower redshifts. With anticipated multi-wavelength spectroscopic data, including from an approved JWST Cycle 1 Program (GO 2566, PI: Khullar), I will help uncover mass assembly pathways in CJ1241 well into the epoch of reionization. Finally, I describe efforts to create efficient machine learning-based frameworks — specifically using simulation-based inference (SBI) — to calculate posterior distributions of key galaxy parameters, and motivate future efforts to study stellar mass assembly in massive galaxies across different epochs.